apple/LensVLM-9B
Full fine-tune vs Qwen/Qwen3.5-9B. Fine-tune: 96% of the weights changed, across 32 of 32 layers.
Tensors changed: lm_head 1/1, Embeddings 2/4, MLP 100/208, Attention 202/332, Norms 107/215
8 derivatives checked
- bartowski/LensVLM-9B-GGUF: Same weights, converted
- prithivMLmods/LensVLM-9B-GGUF: Same weights, converted
- mradermacher/LensVLM-9B-i1-GGUF: Quantization
- mradermacher/LensVLM-9B-GGUF: Same weights, converted
- mlx-community/LensVLM-9B-OptiQ-4bit: Quantization
- prithivMLmods/LensVLM-9B-MLX: Partial fine-tune
- sahilchachra/LensVLM-9B-MXFP4: Quantization
- sahilchachra/LensVLM-9B-MXFP8: Quantization
2 other repos hold byte-for-byte identical weight files.